Voice based speaker recognition has been anticipated as a convenient method for establishing a user's identity (authentication) in mobile terminals since it may be implemented without additional hardware and can therefore be applied to existing terminals. Unfortunately, voice based authentication suffers from poor accuracy as compared with other biometric authentication methods, i.e., fingerprint, retinal scan.